pick [opt](merge-on-write) avoid to check delete bitmap while lookup …

…rowkey in some situation to reduce CPU cost (apache#41480) (apache#41439) Issue Number: close #xxx cherry-pick apache#41480
bobhan1 · Oct 15, 2024 · ccf869c · ccf869c
1 parent fe069fd
commit ccf869c
Show file tree

Hide file tree

Showing 4 changed files with 22 additions and 15 deletions.
diff --git a/be/src/olap/rowset/segment_v2/segment_writer.cpp b/be/src/olap/rowset/segment_v2/segment_writer.cpp
@@ -363,7 +363,7 @@ Status SegmentWriter::probe_key_for_mow(
     RowsetSharedPtr rowset;
     auto st = tablet->lookup_row_key(key, _tablet_schema.get(), have_input_seq_column,
                                      specified_rowsets, &loc, _mow_context->max_version,
-                                     segment_caches, &rowset);
+                                     segment_caches, &rowset, true, true);
     if (st.is<KEY_NOT_FOUND>()) {
         if (_opts.rowset_ctx->partial_update_info->is_strict_mode) {
             ++stats.num_rows_filtered;
@@ -867,7 +867,7 @@ Status SegmentWriter::merge_rows_for_sequence_column(
 
             st = tablet->lookup_row_key(key, _tablet_schema.get(), false, specified_rowsets, &loc,
                                         _mow_context->max_version, segment_caches, &rowset, true,
-                                        &previous_encoded_seq_value);
+                                        true, &previous_encoded_seq_value);
             DCHECK(st.is<KEY_NOT_FOUND>() || st.ok());
 
             Slice previous_seq_slice {};

diff --git a/be/src/olap/rowset/segment_v2/vertical_segment_writer.cpp b/be/src/olap/rowset/segment_v2/vertical_segment_writer.cpp
@@ -316,7 +316,7 @@ Status VerticalSegmentWriter::_probe_key_for_mow(
     RowsetSharedPtr rowset;
     auto st = tablet->lookup_row_key(key, _tablet_schema.get(), have_input_seq_column,
                                      specified_rowsets, &loc, _mow_context->max_version,
-                                     segment_caches, &rowset);
+                                     segment_caches, &rowset, true, true);
     if (st.is<KEY_NOT_FOUND>()) {
         if (_opts.rowset_ctx->partial_update_info->is_strict_mode) {
             ++stats.num_rows_filtered;
@@ -860,7 +860,7 @@ Status VerticalSegmentWriter::_merge_rows_for_sequence_column(
             std::string previous_encoded_seq_value {};
             st = tablet->lookup_row_key(key, _tablet_schema.get(), false, specified_rowsets, &loc,
                                         _mow_context->max_version, segment_caches, &rowset, true,
-                                        &previous_encoded_seq_value);
+                                        true, &previous_encoded_seq_value);
             DCHECK(st.is<KEY_NOT_FOUND>() || st.ok());
 
             Slice previous_seq_slice {};

diff --git a/be/src/olap/tablet.cpp b/be/src/olap/tablet.cpp
@@ -2856,7 +2856,7 @@ Status Tablet::lookup_row_key(const Slice& encoded_key, TabletSchema* latest_sch
                               const std::vector<RowsetSharedPtr>& specified_rowsets,
                               RowLocation* row_location, uint32_t version,
                               std::vector<std::unique_ptr<SegmentCacheHandle>>& segment_caches,
-                              RowsetSharedPtr* rowset, bool with_rowid,
+                              RowsetSharedPtr* rowset, bool with_rowid, bool is_partial_update,
                               std::string* encoded_seq_value) {
     SCOPED_BVAR_LATENCY(g_tablet_lookup_rowkey_latency);
     size_t seq_col_length = 0;
@@ -2874,6 +2874,8 @@ Status Tablet::lookup_row_key(const Slice& encoded_key, TabletSchema* latest_sch
             Slice(encoded_key.get_data(), encoded_key.get_size() - seq_col_length - rowid_length);
     RowLocation loc;
 
+    bool need_to_check_delete_bitmap = is_partial_update || with_seq_col;
+
     for (size_t i = 0; i < specified_rowsets.size(); i++) {
         auto& rs = specified_rowsets[i];
         auto& segments_key_bounds = rs->rowset_meta()->get_segments_key_bounds();
@@ -2912,15 +2914,19 @@ Status Tablet::lookup_row_key(const Slice& encoded_key, TabletSchema* latest_sch
             if (!s.ok() && !s.is<KEY_ALREADY_EXISTS>()) {
                 return s;
             }
-            if (s.ok() && _tablet_meta->delete_bitmap().contains_agg_without_cache(
-                                  {loc.rowset_id, loc.segment_id, version}, loc.row_id)) {
-                // if has sequence col, we continue to compare the sequence_id of
-                // all rowsets, util we find an existing key.
-                if (schema->has_sequence_col()) {
-                    continue;
+            if (s.ok() && need_to_check_delete_bitmap) {
+                // check if the key is already mark deleted
+                if (_tablet_meta->delete_bitmap().contains_agg_without_cache(
+                            {loc.rowset_id, loc.segment_id, version}, loc.row_id)) {
+                    // if has sequence col, we continue to compare the sequence_id of
+                    // all rowsets, util we find an existing key.
+                    if (with_seq_col) {
+                        continue;
+                    }
+                    // The key is deleted, we need to break the loop and return
+                    // KEY_NOT_FOUND.
+                    break;
                 }
-                // The key is deleted, we don't need to search for it any more.
-                break;
             }
             // `st` is either OK or KEY_ALREADY_EXISTS now.
             // for partial update, even if the key is already exists, we still need to
@@ -3097,7 +3103,8 @@ Status Tablet::calc_segment_delete_bitmap(RowsetSharedPtr rowset,
 
             RowsetSharedPtr rowset_find;
             auto st = lookup_row_key(key, rowset_schema.get(), true, specified_rowsets, &loc,
-                                     dummy_version.first - 1, segment_caches, &rowset_find);
+                                     dummy_version.first - 1, segment_caches, &rowset_find, false,
+                                     is_partial_update);
             bool expected_st = st.ok() || st.is<KEY_NOT_FOUND>() || st.is<KEY_ALREADY_EXISTS>();
             // It's a defensive DCHECK, we need to exclude some common errors to avoid core-dump
             // while stress test

diff --git a/be/src/olap/tablet.h b/be/src/olap/tablet.h
@@ -439,7 +439,7 @@ class Tablet final : public BaseTablet {
                           RowLocation* row_location, uint32_t version,
                           std::vector<std::unique_ptr<SegmentCacheHandle>>& segment_caches,
                           RowsetSharedPtr* rowset = nullptr, bool with_rowid = true,
-                          std::string* encoded_seq_value = nullptr);
+                          bool is_partial_update = false, std::string* encoded_seq_value = nullptr);
 
     // Lookup a row with TupleDescriptor and fill Block
     Status lookup_row_data(const Slice& encoded_key, const RowLocation& row_location,